Moderator: our reporter Huang Wei

Dialogue experts: Professor Xiong Zhang, School of Computer Science, Beihang University, Associate Professor Wang Jingyuan

Dialogue background: Ethical education in artificial intelligence education has two aspects: one is the ethics of artificial intelligence system design goals, technical routes, and function realization. The value orientation is to promote scientific and technological progress, promote social development, and improve people's lives, and must not violate the welfare of the public; the other is the ethics when using artificial intelligence systems. Simply put, it is to promote good and avoid evil, and do things that benefit mankind, and cannot use artificial intelligence technology to do things that harm society and others. The reporter recently had a conversation with two professors, Xiong Zhang and Wang Jingyuan, from the School of Computer Science, Beihang University, on the topic of artificial intelligence and ethics education.

To ensure healthy artificial intelligence education, we must pay attention to its ethical education

Moderator: Using machines to replace human brain work has been a dream of mankind since the invention of computers. After decades of hard work, artificial intelligence technology has experienced several ups and downs, and has become increasingly mature. It has begun to assist or replace some of human mental work. In some tasks with certain goals (such as Go, face recognition, etc.), artificial intelligence can even surpass humans and show better performance. In artificial intelligence education, we must pay attention to its scientific principles, development laws, and impact on social development. So where should we place ethical education in artificial intelligence education?

Xiong Zhang: Sickles and axes, steam engines, electricity, and telephones have extended and enhanced human limbs and facial features. Only artificial intelligence has extended and enhanced human brainpower for the first time. The biggest difference between artificial intelligence technology and all previous technological advances is that it can liberate humans from some mental work. As artificial intelligence technology continues to mature and be widely used, more and more judgments and decisions will be completed by artificial intelligence in the future, such as identity confirmation, homework evaluation, reading medical records, resume screening, investment decisions, etc. This means that artificial intelligence will be deeply involved in many important activities of human society.

We have to admit that when humans make judgments and decisions, their basis must include socially recognized norms and the emotions, attitudes, and values ​​of the decision-maker. Society has many ethical norms and constraints on human decision-makers. When a machine participates in human judgment and decision-making and plays the role of artificial intelligence, it is easy for the system to judge based on a set of objective indicators. When the system is engaged in decision-making affairs, it is a major challenge for artificial intelligence to be able to obey socially recognized norms and human values.

Wang Jingyuan: For example, a reasonable artificial intelligence resume screening software should not conduct discriminatory resume screening based on factors such as gender, race, skin color, etc.; a software used for smart medical care should not exaggerate or cover up the condition in order to obtain improper benefits, or even delay treatment. Ethical and moral standards that require the behavior of human decision-makers, such as fairness, justice, non-discrimination, and non-deception, also need to be effectively reflected in the automatic decision-making of artificial intelligence. An important task of education is to help people establish correct values ​​and moral standards such as kindness, integrity, fairness, and justice. With the current level of artificial intelligence technology, it is still impossible for us to provide moral education for artificial intelligence programs, but when we carry out artificial intelligence education, we must not ignore ethical education in artificial intelligence education.

Several basic ethical education points in artificial intelligence education

Moderator: Educational developers should fully consider the basic requirements of social ethics when designing and developing artificial intelligence tools. In addition to complying with the basic principles of non-harmful ethics of information technology, such as not creating computer viruses, not developing harmful programs based on artificial intelligence, and not engaging in data theft, are there any other considerations?

Xiong Zhang: We must also fully consider the new role of artificial intelligence in social activities, take into account the ethical risks that may arise from the inappropriate behavior of these roles, and refrain from actively using the special status of the role for personal gain or harming the interests of others.

For example, when developers develop a human resources management system for a company and use traditional databases, information systems and other technologies, ethical developers need to do not use their own technical advantages to reserve backdoors in the program, cannot tamper with or copy the company's internal data, and fully consider information security factors when designing programs. When developing a human resources management system based on artificial intelligence, developers also need to consider the ethical and moral risks of machines replacing humans to do evil. For example, we should proactively avoid inputting factors such as race, skin color, gender, etc. as features into automatic decision-making algorithms, and should not develop algorithm functions that pry into employees' personal privacy.

Furthermore, artificial intelligence ethics education must also teach users to learn to use technology correctly in reasonable scenarios. Technology is usually neutral, and good and evil are only reflected in the usage scenarios of the technology. For example, the video "face-changing" technology based on artificial intelligence was originally invented to be used in film special effects, artistic creation, etc., but it may also be used in inappropriate scenarios such as bad publicity and falsified evidence.

In addition, passive ethical risks of artificial intelligence must be avoided. This means that when artificial intelligence performs important work functions, it may be interfered by uncertain factors such as environmental changes and external attacks. If the artificial intelligence technology used in important positions is not reliable and robust enough, it may have some unpredictable consequences when attacked by the outside world. For example, during autonomous driving, special lighting and other scenarios may cause the algorithm to turn a blind eye to obvious obstacles. For image recognition algorithms in "face-swiping" payment applications, some adversarial samples can completely confuse the output of the artificial intelligence model and open the account to visitors who should not be authorized. These are all risk issues caused by insufficient reliability of artificial intelligence in key positions.

Wang Jingyuan: It is necessary to educate students to proactively understand the limitations, uncertainties and risk consequences of technology when using artificial intelligence technology to replace human work in the future.

For the entire society, both developers and users of artificial intelligence technology need to carry out necessary responsibility awareness education. The important feature of artificial intelligence tools is to assist and act on behalf of humans to complete certain mental tasks. When machines complete certain tasks on behalf of humans, "agency risk" will inevitably arise.

For example, in traditional warfare, weapon operators need to go to the battlefield to fight in person, so they have an intuitive understanding of the cruelty of killing. In the era of unmanned weapons, weapon operators only need to click a button in the office to kill enemies thousands of miles away. The cruelty of war has been reduced to the level of video games for operators.

In the era of artificial intelligence, weapons killing may be completely determined by algorithms, and people's life and death are completely controlled by algorithms. In this process, if there are any unexpected impacts due to the unreliability of artificial intelligence, the consequences may even bring huge disasters to the entire human race. To avoid such a tragedy, necessary technical control measures are indispensable. It is also very important to strengthen the responsibility awareness education for the development and users of artificial intelligence tools.

Management departments’ actions in promoting the healthy development of artificial intelligence education

Moderator: Artificial intelligence is a developing science and technology. In its development process, in addition to promoting science and technology and promoting social development, the integration of artificial intelligence and education has also given birth to accompanying evaluation, personalized guidance, teamwork, and innovated the teaching environment, atmosphere and culture. New changes and new impacts will appear. What should our schools and management departments do?

Xiong Zhang: Ethical education in artificial intelligence education is a process of continuous development and improvement. There is a long way to go to maintain the scientific nature of this process and maintain the healthy development of artificial intelligence education. Artificial intelligence education should be incorporated into the curriculum system of compulsory education, ordinary high schools and universities to ensure that schools serve as the main channel for artificial intelligence education. The education management department must make it clear in the curriculum plans of ordinary high schools and compulsory education that artificial intelligence courses are an important part of information technology courses and information science courses, clarify training objectives, course content and academic quality standards, and constantly update them in a timely manner. They must strictly ensure the scientificity, rigor and standardization of school textbooks, and train front-line teachers in schools in a timely manner so that they can become the main force. At the same time, public opinion and social forces must be guided to jointly safeguard the healthy development of artificial intelligence education.

"China Education News" June 15, 2019 Page 3